Can Information Flows Suggest Targets for Interventions in Neural Circuits?
arXiv:2111.05299
Abstract
Motivated by neuroscientific and clinical applications, we empirically examine whether observational measures of information flow can suggest interventions. We do so by performing experiments on artificial neural networks in the context of fairness in machine learning, where the goal is to induce fairness in the system through interventions. Using our recently developed -information flow framework, we measure the flow of information about the true label (responsible for accuracy, and hence desirable), and separately, the flow of information about a protected attribute (responsible for bias, and hence undesirable) on the edges of a trained neural network. We then compare the flow magnitudes against the effect of intervening on those edges by pruning. We show that pruning edges that carry larger information flows about the protected attribute reduces bias at the output to a greater extent. This demonstrates that -information flow can meaningfully suggest targets for interventions, answering the title's question in the affirmative. We also evaluate bias-accuracy tradeoffs for different intervention strategies, to analyze how one might use estimates of desirable and undesirable information flows (here, accuracy and bias flows) to inform interventions that preserve the former while reducing the latter.
Accepted to the Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS 2021). (29 pages; 61 figures)
References in corpus (5)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Is There a Trade-Off Between Fairness and Accuracy? A Perspective Using Mismatched Hypothesis Testing
- FACT: A Diagnostic for Group Fairness Trade-offs
- Recovering from Biased Data: Can Fairness Constraints Improve Accuracy?
- To Split or Not to Split: The Impact of Disparate Treatment in Classification